AI & Amo.ng Definition of Terms

A

Data and privacy

Access Control

Rules and mechanisms that determine who or what can view or change a resource. Giving a tool access should be limited to what the authorised task requires.

AI agents and workflows

Software that can use a model to choose and carry out steps toward a goal, often with tools and feedback. Its actual autonomy depends on its design and permissions; the label does not mean it should act without human oversight.

Amo.ng

Agent Skill package

A downloadable package offered for selected Amo.ng Skills, containing portable instructions and supporting source material for compatible agent tools. Downloading does not install or execute it, grant tool permissions, or add verification beyond the source’s existing status.

AI agents and workflows

Agentic AI

A broad term for AI systems that plan or take multiple actions toward a goal. It is not a uniform product standard: check what the system can actually access, change and do without approval.

Quality, evidence and governance

AI Governance

The roles, policies, decisions and controls used to oversee AI through its lifecycle. Governance includes responsibility for use, evidence, monitoring, risk and human oversight; a policy document alone is not enough.

Development and integrations

AI Provider

An organisation or service supplying access to AI models. In Workplace, you choose a supported provider for generation using your own API access. Provider availability, charges, limits and data terms are separate from Amo.ng’s free library.

Amo.ng

Amo ID

A stable identifier for a resource, such as AMO-P-000155 for a Prompt, AMO-W-000044 for a Workflow or AMO-S-000007 for a Skill. The ID identifies the resource, not a particular edition of its content.

Amo.ng

Amo Verified

A public verification label tied to a particular published Prompt version and its recorded review decision. Read any stated limitations. It is not a promise that every generated result is correct or that using the Prompt establishes professional compliance.

Amo.ng

A free library and toolkit for practical AI work: discover Prompts, Workflows and Skills, explore Professional Packs, work in Workplace and create supported websites or browser tools with Amo.ng Build. Amo.ng is not a foundation model.

Amo.ng

Amo.ng Build

The AI-assisted builder for supported websites and browser tools. You can start with an idea or compatible Prompt, review the brief, generate and revise a project, then save, download or publish when ready. Human testing remains necessary.

Data and privacy

Anonymisation

Changing information so people are no longer identifiable, taking realistic re-identification risks into account. Removing names alone may be insufficient. Applicable standards for judging anonymity depend on the context and law.

Development and integrations

API

An Application Programming Interface: an agreed way for software to request information or operations from other software. Amo.ng’s public API retrieves published resources; it does not run them or change library content.

Development and integrations

API Credit / Usage

API credits are a service’s balance or allowance for API consumption. Usage is the amount consumed, which may be measured in tokens, images, time or requests. Prices and expiry rules depend on the provider; Amo.ng’s free access does not include these charges.

Data and privacy

API Key

A credential used to access a service’s API and often associate usage or billing with an account. Keep it out of Prompts, screenshots and shared documents. In Workplace, use the designated provider-key controls.

Quality, evidence and governance

Approval

An authorised person’s decision to permit a defined use or action. A recorded approval needs an appropriate scope and authority; an AI recommendation or a button click does not create professional authority.

AI fundamentals

Artificial Intelligence / AI

Computer systems that use inputs to produce predictions, recommendations, content or other outputs for a task. AI systems vary in capability and autonomy; an AI label does not imply human understanding or reliable judgement.

Amo.ng

Assistant

On Amo.ng, the Assistant recommends relevant Prompts, an order of use and review steps from the outcome you describe. It does not execute the recommended tasks. Elsewhere, “AI assistant” can mean a conversational tool with different capabilities.

Quality, evidence and governance

Assumption

Something provisionally treated as true to proceed with analysis. Label assumptions and test the important ones rather than presenting them as supplied facts.

B

Quality, evidence and governance

Benchmark

A defined test or reference used to compare systems. A high benchmark result may not predict performance on your own data, users or professional task.

Quality, evidence and governance

Bias

A systematic tendency that can distort data, analysis or decisions. In AI, it may arise from data, design or use and can disadvantage people. A statistical bias and unfair treatment are related but not identical concepts.

Amo.ng

Bring Your Own AI / BYOAI

Using your own supported AI provider and API access for generation. Amo.ng access does not include free provider usage; your provider’s charges and data-handling terms apply. A consumer chat subscription may not include API access.

Amo.ng

Build & Ship

An Amo.ng discovery area for Prompts and Workflows that help plan, implement, test and prepare software for release. It is distinct from Amo.ng Build, the building application. A Build & Ship Prompt is not automatically compatible with Build.

Amo.ng

Build-compatible Prompt

A published Prompt that Amo.ng Build recognises as a supported starting point for a project. Importing its instructions does not build or publish anything by itself. Review the inputs and project brief before generation.

C

Amo.ng

Carry-forward

The reviewed output or context taken from one Workflow step into the next. Carry forward only the relevant material, preserving qualifications, missing evidence and human decisions rather than treating every generated statement as fact.

Quality, evidence and governance

Causal Claim

A statement that one factor caused a change in another. It needs a defensible design or reasoning that addresses alternative explanations; seeing two measures move together is not enough.

Quality, evidence and governance

Citation

A reference identifying material used to support a statement. Open and check it: a citation can be mistaken, irrelevant or fabricated, and a real source may not support the claim beside it.

Quality, evidence and governance

Claim

A statement presented for acceptance as true. Identify its source, scope and supporting evidence, especially before publishing or making a consequential decision.

Development and integrations

Cloud Model

A model accessed on a provider’s remote infrastructure. Inputs and outputs are subject to the service’s processing, retention and access arrangements, even if the app on your device feels private.

Quality, evidence and governance

Confidence

The degree of certainty expressed about a result. A model’s confident wording or self-assigned score is not automatically a calibrated probability. Statistical confidence has a more specific meaning tied to a stated method.

Development and integrations

Connector

An integration that connects an application to another data source or service. Check what it can read or change and whose permissions it uses.

Quality, evidence and governance

Constraint

A limit or condition the work must respect, such as an approved scope, deadline, budget or data-use restriction. State constraints before using AI so proposed outputs can be checked against them.

AI fundamentals

Context

Information available to a model for a response, such as instructions, conversation history and supplied documents. The model cannot safely be assumed to know material that has not been provided or retrieved.

AI fundamentals

Context Window

The amount of information a model can handle in one interaction, usually measured in tokens. Depending on the service, input, history, tool results and generated output share limits; a long conversation may be shortened or summarised.

Amo.ng

Copy Prompt

Copies the Prompt instructions so you can paste them into an AI tool. Copying does not execute the Prompt or send it to an AI provider. You still need to supply the required inputs and run it in your chosen tool.

Amo.ng

Copy Skill

Copies the Skill’s purpose, required inputs, usage instructions, source reference and completion criteria. It does not automatically copy the entire underlying source Prompt or Workflow, install a package or execute the Skill.

Amo.ng

Copy Workflow

Copies the Workflow outcome, required inputs, ordered steps, review and carry-forward guidance, completion criteria, and the publicly available linked Prompt instructions. It copies text; it does not start a run.

Quality, evidence and governance

Correlation

A statistical relationship in how quantities vary together. It can suggest a question to investigate, but does not by itself show that one quantity caused the other to change.

Data and privacy

Credential

Information or a mechanism used to establish an identity or access right, such as a password, token or API key. Treat credentials separately from the content being analysed.

D

Data and privacy

Data Deletion

Removing information from a system. Deleting one copy does not necessarily remove backups, exports or copies already sent elsewhere; understand the relevant product and provider’s deletion process.

Quality, evidence and governance

Data Drift

A change in the data a system receives compared with earlier or training data. Drift can signal that past evaluation results may no longer represent current use.

Data and privacy

Data Minimisation

Using only the information needed for a stated purpose. For AI work, remove unnecessary fields and passages before sending material to a tool, while retaining enough context for a sound result.

Data and privacy

Data Retention

Keeping information for a specified purpose and period. Agree retention based on the need and applicable obligations; using AI does not create a universal retention period.

AI fundamentals

Deep Learning

Machine learning using neural networks with many processing layers. These layers learn different representations of data; “deep” describes the architecture, not the depth of human-like understanding.

AI fundamentals

Deterministic

Producing the same result when the inputs and relevant conditions are the same. AI services may still vary despite settings intended to reduce randomness, so do not assume a response can always be reproduced exactly.

E

AI fundamentals

Embedding

A numerical representation of an item, such as a passage or image, used to compare patterns or meaning. Similar embeddings can help find related material; they are not a human-readable explanation of it.

Data and privacy

Encryption

Transforming information into a form that requires an appropriate key to read. It protects data under specified conditions, but does not prevent misuse by an authorised recipient or a compromised unlocked device.

Data and privacy

Encryption at Rest

Encryption protecting stored information, such as a file or local vault. Once unlocked, the information can be available to the application and anyone controlling that session.

Data and privacy

Encryption in Transit

Encryption protecting information while it travels between systems, commonly through HTTPS. The receiving system can still process the information; secure transport is not permission to disclose it.

Quality, evidence and governance

Evaluation / Eval

A planned assessment of performance against defined tasks and criteria. Useful evaluations include realistic examples, failures and limits, rather than relying on one impressive response.

Quality, evidence and governance

Evidence

Information that supports or challenges a claim in a particular context. Its relevance, reliability and sufficiency need assessment. An AI response or Run Record is not automatically evidence that the underlying facts are correct.

Quality, evidence and governance

Explainability

The ability to provide an understandable account of a system’s behaviour or result. An explanation can help review, but a model-generated explanation is not automatically a faithful account of its internal process.

F

Quality, evidence and governance

Fairness

An assessment of whether a system treats people and groups appropriately in its context. Different fairness goals can conflict, and a single metric does not settle ethical, professional or legal questions.

AI fundamentals

Fine-tuning

Further training an existing model on selected examples to change how it performs or responds. This changes model parameters; supplying reference documents in a request does not.

Quality, evidence and governance

Forecast

An estimate of future outcomes based on current information, assumptions and expected conditions. A forecast can change as evidence changes; it is not the same as a target or commitment.

AI fundamentals

Foundation Model

A model trained on broad data that can be adapted or used for many tasks. It still has limits, and suitability must be checked for the intended use. Applications can build on foundation models without being models themselves.

AI agents and workflows

Function Calling / Function Call

A mechanism through which a model proposes a named software operation and its arguments in a structured form. The surrounding application decides whether and how to execute it; a proposed call is not proof that an action succeeded.

G

AI fundamentals

Generative AI

AI that produces content such as text, images, audio or code based on learned patterns and supplied instructions. A fluent or realistic result may still be inaccurate.

AI fundamentals

Grounding

Connecting a response to supplied or retrieved information. Grounding makes claims easier to check but does not ensure that the sources are reliable or that the model used them correctly.

AI agents and workflows

Guardrail

A control intended to limit unwanted system behaviour, such as input checks, tool permissions or approval requirements. Guardrails reduce certain risks; they are not a guarantee that a system cannot fail.

H

AI fundamentals

Hallucination

An AI response that presents invented, incorrect or unsupported content as if it were sound. Examples include fabricated citations or details absent from supplied evidence. Confidence of wording is not a reliability test.

Quality, evidence and governance

Human Review

A person examines evidence, assumptions, limitations and output before deciding what can be relied on. In Workplace, recording review or approval documents a user action; it does not prove that the underlying work is correct.

AI agents and workflows

Human-in-the-Loop

A process with specified points where a person checks, decides or intervenes. It is meaningful only when the person has enough information, authority and time to challenge the result.

Quality, evidence and governance

Hypothesis

A proposed explanation or expectation that can be examined against evidence. Keep it separate from established findings and specify what observations could challenge it.

I

AI fundamentals

Inference

Using an already trained model to produce an output from new input. Generating a response is inference, even when the product describes part of the process as reasoning.

Development and integrations

Integration

A connection that lets systems exchange information or work together. Amo.ng integrations can help other clients discover and retrieve resources; the connection does not automatically execute those resources.

Quality, evidence and governance

Interpretability

How readily people can understand a model or the meaning of its output in context. The term overlaps with explainability; make clear which kind of understanding is required for the task.

J

AI agents and workflows

Jailbreak

An attempt to get a model or application to bypass its intended restrictions. The term overlaps with some prompt-injection attacks, but not every misleading instruction has the same purpose or entry point.

Development and integrations

JSON

JavaScript Object Notation: a text format for exchanging values in named fields and lists. It is widely used by APIs. Correct JSON syntax says nothing about whether its contents are true.

L

AI fundamentals

Large Language Model / LLM

A large model trained to process and generate language. It can draft, summarise and analyse text, but its ability to produce a plausible answer does not establish that the answer is true.

Development and integrations

Latency

The delay before a system responds. Time to the first visible output and time to complete the whole response are different measures.

Development and integrations

Local Model

A model run on a device or infrastructure you control instead of a hosted model endpoint. Local inference does not automatically make every connected application, download or telemetry process offline.

M

AI fundamentals

Machine Learning

Methods that learn patterns from examples or experience instead of relying entirely on hand-written rules for each situation. Performance depends on the data, objective and conditions in which the system is used.

Amo.ng

Manual / External step

Work performed by a person or in another system, such as an interview, a review or a Codex task. Recording an outcome in Workplace does not mean Workplace performed or independently verified that work.

Development and integrations

Markdown

A plain-text format using simple markers for headings, lists, links and tables. Amo.ng uses it for some editable downloads and portable instructions. It is not a PDF, Word file or spreadsheet.

Development and integrations

MCP / Model Context Protocol

An open protocol through which AI applications connect to external information and tools. Amo.ng’s MCP server provides read-only resource discovery and retrieval. Other MCP servers may offer actions, so check each server’s permissions and purpose.

AI fundamentals

Model

A learned or designed computational representation used to turn inputs into outputs. In an AI service, the model is the underlying capability; the provider or application supplies the surrounding service.

Quality, evidence and governance

Model Drift

A change or deterioration in a model’s behaviour or suitability over time, such as when relationships in real-world data change. People use the term differently; record which performance measure or relationship changed.

Development and integrations

Model Endpoint

A service address that accepts requests to a model or model service. Access normally depends on credentials, supported request formats and provider limits.

Quality, evidence and governance

Model Evaluation

Testing a model’s capabilities and limitations for a specified purpose. Model tests should be complemented by checks of the whole application, including retrieval, tools and human review.

AI fundamentals

Multimodal Model

A model that handles more than one kind of information, such as text and images or audio and text. Supported input and output types vary by model.

N

AI fundamentals

Neural Network

A model made of connected computational units whose adjustable weights transform inputs into outputs. The brain-inspired name does not mean it works like a person’s brain.

O

Quality, evidence and governance

Observation

Something directly recorded or measured, with its source and conditions. Separate it from an explanation of why it happened or a prediction about what happens next.

Amo.ng

Open resource

Opens the normal Amo.ng detail page so you can read a Prompt, Workflow or Skill and its usage guidance. It does not run the resource, send task inputs to a provider or download a file.

Development and integrations

Open-source Model

A model offered with the freedoms and materials needed to use, study, modify and share it under the applicable open-source definition and licence. Check the actual release; downloadable weights alone do not establish this status.

Development and integrations

Open-weight Model

A model whose trained parameters are made available for others to download under stated terms. Those terms may restrict use, and weights alone do not supply all the information and permissions required for open-source AI.

AI fundamentals

Output

What a system produces, such as an answer, image, code or proposed action. AI-generated output is working material until checked for its intended use.

P

Data and privacy

Permission

Authorisation to access information or perform an action within a defined scope. Technical access, organisational approval and a person’s consent are different things; having one does not automatically supply the others.

Data and privacy

Personal Data / Personal Information

Information that identifies, or can reasonably be linked to, a person. Names are only one example; combinations of records can also identify someone. Exact legal definitions depend on jurisdiction.

Data and privacy

Personally Identifiable Information / PII

A term commonly used for information that identifies or can be linked to an individual. Its scope varies across standards and laws; do not assume that data outside a narrow PII list is safe to disclose.

Development and integrations

Plugin

An add-on that extends an application’s capabilities. Different products use the word differently; installing a plugin may introduce new access or actions that need review.

Amo.ng

Practice Case

Material used to practise analysis and judgement. Check whether it is fictional or based on a real event, and keep learner evidence separate from any facilitator guidance. Completing a case does not confer professional certification.

Amo.ng

Private Results

The Workplace library for results you explicitly save in this browser and reopen after unlocking your vault. Clearing site data may remove them. A vault backup protects saved keys; Private Results need their own downloads.

Quality, evidence and governance

Professional Judgement

Applying relevant knowledge, experience, evidence and responsibilities to a decision in context. AI can help organise working material, but it does not take over the accountable professional’s judgement or authority.

Amo.ng

Professional Pack

A curated collection of Amo.ng resources designed around the real work of a profession, business function or field. Packs organise resources into useful objectives and journeys while keeping professional judgement with the user.

Amo.ng

Prompt

Instructions and context given to an AI tool. On Amo.ng, a Prompt is a structured set of instructions for completing one task, with guidance on the inputs to provide and how to use the result.

AI fundamentals

Prompt Engineering

Designing and testing instructions, context and examples to improve AI responses for a task. Better prompts can help, but cannot guarantee factual accuracy or replace missing evidence.

AI agents and workflows

Prompt Injection

An attempt to make an AI application follow untrusted instructions, for example inside a retrieved webpage or document, instead of its authorised task. Treating external content as evidence rather than authority helps limit this risk.

Quality, evidence and governance

Provenance

Information about where material came from and how it was created or changed. A provenance record supports investigation; it does not by itself prove the material is correct.

Amo.ng

Provider-supported step

A step that the current Workplace execution support can send to a supported AI provider. Availability depends on the resource, tool and configured support. It does not authorise spending, external actions or professional decisions.

Data and privacy

Pseudonymisation

Replacing identifying details with labels while keeping information that could reconnect them to people separately. It reduces direct identification but is not the same as anonymisation.

R

Development and integrations

Rate Limit

A limit on how much work a service accepts within a period, such as requests or tokens per minute. It is different from the total money or credits available in an account.

Amo.ng

Real Case Study

A learning account of a real event supported by identified sources. It should distinguish allegations from findings and explain the relevant decision, settlement or appeal status. Lessons do not automatically establish the same conclusion in another case.

Quality, evidence and governance

Reconciliation

Comparing two sets of records or balances, identifying differences and explaining or resolving them with evidence. Matching totals alone does not prove that the records are correct.

Data and privacy

Redaction

Removing or concealing sensitive content before sharing. Check the actual exported file: a visual cover or changed display may leave the underlying information recoverable.

Quality, evidence and governance

Responsible AI

Developing and using AI with attention to people, accountability, safety, fairness and the consequences of its use. It requires ongoing practice rather than a label that proves a system is trustworthy.

AI fundamentals

Retrieval-Augmented Generation / RAG

A method that retrieves relevant material and supplies it to a model to help generate an answer. It can use current or private documents without retraining, but retrieval and answer accuracy still need checking.

Amo.ng

Run in Workplace

Opens the private Amo.ng workspace for a supported Prompt or Workflow. You supply inputs and select your own supported AI provider before generation. Opening the workspace alone does not generate output; provider API charges may apply.

Amo.ng

Run Record

A record of Workflow activity, outputs and human decisions recorded in Workplace. It can help someone review what was recorded, but does not independently prove the work occurred correctly or establish audit evidence, certification or professional approval.

S

AI agents and workflows

Safety Filter

A mechanism that detects or blocks content or actions under a service’s safety rules. It can make mistakes and does not replace professional review or access controls.

Development and integrations

Schema

A description of the expected structure and types of data, including required fields. Schema checks can catch missing or malformed values; they cannot establish professional validity.

Development and integrations

SDK

A Software Development Kit: libraries and supporting tools that help developers use a service or platform. It simplifies integration but does not remove the service’s authentication, permission or usage requirements.

Data and privacy

Secret

Sensitive access material such as a password, API key or signing key that must be protected from unauthorised disclosure. It should not be embedded in public code or supplied as ordinary task evidence.

AI fundamentals

Semantic Search

Finding material by intended meaning rather than only exact matching words. It can retrieve related concepts, but may also miss important distinctions in the user’s question.

Data and privacy

Sensitive Data

Information whose disclosure or misuse could cause particular harm, such as health details, credentials or confidential business information. Legal categories of sensitive personal data differ by jurisdiction.

Quality, evidence and governance

Sensitivity Analysis

Testing how a result changes when one or more assumptions change, to see which assumptions matter most. The result depends on the model and inputs; it does not prove that a scenario will occur.

Amo.ng

Session-only

Working without saving to a remembered Workplace vault. Session-only provider keys are held for the current page session and disappear on closing or reloading it. This does not erase downloaded files or information already sent to a provider.

Amo.ng

Skill / AI Skill

Reusable instructions for applying a focused capability. An Amo.ng Skill can be based on a Prompt or Workflow and includes usage guidance. It does not automatically run itself. Other AI products may use “skill” for their own package formats.

Quality, evidence and governance

Source

The origin of information, such as a document, person, dataset or system. Check whether it is primary, current and relevant. On Amo.ng, a source resource can also mean the Prompt or Workflow underlying another resource.

Amo.ng

Source Prompt

The underlying published Prompt referenced by a Workflow step or Prompt-based Skill. It supplies task instructions. A link to it is not the same as executing it or copying its entire body.

Amo.ng

Start Here

A recommended entry point for a particular objective or Pack. It helps a newcomer begin; it is not a requirement to complete every other resource or a guarantee that one journey suits every situation.

AI fundamentals

Stochastic

Involving randomness or probability. A stochastic generation process can produce different responses to the same request.

Development and integrations

Streaming

Delivering a response in pieces as it is generated, instead of waiting for the full result. Visible text may still be incomplete or interrupted; review the final output before using it.

Development and integrations

Structured Data

Information organised into consistent fields, records or relationships so software can work with it. Tables and JSON objects are common examples. Structure does not guarantee data quality.

Development and integrations

Structured Output

Output arranged in a specified format, such as JSON with named fields. Valid structure helps software read it, but does not prove the values are accurate or safe.

Amo.ng

Synthetic Case

A deliberately fictional scenario and evidence set for practice. Its invented organisations, people and records must not be presented as real findings or proof about a real organisation.

AI fundamentals

System Prompt

Instructions set by the application or provider to shape an AI interaction, such as its behaviour and constraints. They are distinct from the user’s task request and are not a complete security control.

T

Amo.ng

Task Specification

Amo.ng’s structured description of a resource for software clients, including its identity, inputs, instructions and relevant source relationships. It describes the task; it is not an execution engine or proof that the task has been completed.

AI fundamentals

Temperature

A generation setting that changes how randomly a model selects possible next tokens. Lower values generally reduce variation; they do not guarantee accuracy, consistency or a fully deterministic response. Support differs by model.

Amo.ng

Template

A structured working document you can fill in or adapt, such as a review sheet, register or decision record. Check the stated download format: a document called a workbook is not necessarily a spreadsheet.

AI fundamentals

Token

A unit a model uses to process information. In text it can be a word, part of a word or punctuation. Token counts affect context limits and often API cost; one token is not always one word.

Development and integrations

Tool

A capability an application can call, such as search, calculation or file access. On Amo.ng resource pages, “tool” can instead name the intended AI product, such as Claude. Neither label alone establishes Workplace execution support.

AI agents and workflows

Tool Use

An AI application’s use of a capability outside the model itself, such as searching documents or running a calculation. The application must control permissions and validate requests and results.

Amo.ng

Topic

A discovery page grouping resources around a subject or practical need. It helps you find relevant material; it is not an execution mode or a professional qualification.

Quality, evidence and governance

Traceability

The ability to follow a result or decision back through its inputs, sources, processing and review. It helps show how a conclusion was reached and where gaps remain.

AI fundamentals

Training

The process of adjusting a model using data and an objective so it can perform tasks. Providing context in an ordinary conversation is not the same as retraining the model; providers’ separate data-use policies still matter.

U

Quality, evidence and governance

Uncertainty

What is not known or cannot be established firmly from available information. Make its source and decision impact explicit rather than hiding it behind a confident conclusion.

Amo.ng

Use in Codex

Prepares task instructions that you can copy or download and take to Codex, where this action is available. Amo.ng does not silently launch Codex or execute the task in the background.

Amo.ng

Use with AI

Copies text to take to your chosen AI tool. For a Workflow, this is a usage guide with steps and source Prompt links, rather than all the Prompt bodies. For a Skill, it is the Skill instructions with a short usage introduction. Nothing runs on click.

AI fundamentals

User Prompt

The instruction, question or other input a user supplies for the immediate task. It may include reference material that the application should treat as data rather than instructions.

V

Quality, evidence and governance

Validation

Checking whether something is suitable for its intended purpose and use. A syntactically valid file or structurally valid response may still fail professional or practical validation.

Amo.ng

Variable

A named placeholder for an input, such as a document or objective, that you supply before running a Prompt. A placeholder is not a fact the AI should invent. On Amo.ng, consult the resource’s input and usage guidance.

Amo.ng

Vault

Optional encrypted storage in your browser for remembered Workplace keys and supported saved work. Its passphrase stays in the browser and has no Amo.ng account recovery. Clearing browser data can remove it; encryption does not protect an unlocked session from someone controlling your device.

AI fundamentals

Vector

An ordered list of numbers. AI systems often use vectors to represent items for comparison, for example when searching document embeddings.

Development and integrations

Vector Database

A system designed to store and search numerical vectors, often alongside ordinary metadata. It can help retrieve similar items, but similarity alone does not establish relevance, permission or truth.

Quality, evidence and governance

Verification

Checking a statement, output or item against specified evidence or requirements. On Amo.ng, the Amo Verified label has its own version-specific review conditions; ordinary checking does not grant that label.

Amo.ng

Version

An identified edition of content or software. A published Amo.ng Prompt version identifies a recorded state of its instructions. Keep the version as well as the Amo ID when the exact content used matters; versioning is not an accuracy guarantee.

W

Development and integrations

Webhook

A notification one system sends to another when an event occurs, usually by making an HTTP request. The receiving system needs to check authenticity and handle repeated or failed notifications.

Amo.ng

Workflow

A step-by-step sequence for completing a larger job, with inputs, review points and handoffs between steps. An Amo.ng Workflow links the Prompts used in that journey; opening or copying it does not run its steps.

AI agents and workflows

Workflow Automation

Software performing a defined sequence of tasks with triggers, conditions and handoffs. It may use AI or ordinary rules. An Amo.ng Workflow is not automatically an unattended automation.

Amo.ng

Workflow Runner

The guided Workplace interface for working through a supported published Workflow. It manages step inputs, results, review decisions and carry-forward material. Manual or external work still needs to be performed and recorded by a person.

Quality, evidence and governance

Working Material

Draft analysis, notes or proposed output that still needs checking before reliance. Treat AI output as working material, even when it looks polished or uses professional language.

Amo.ng

Workplace

The private Amo.ng workspace where supported Prompts and Workflow steps can be run using your own supported AI provider. You choose inputs and review results. Running a request sends the required content to that provider; “private” does not mean offline.

Definition of Terms

Find a term

A

B

C

D

E

F

G

H

I

J

L

M

N

O

P

R

S

T